A proactive learning engine that enables AI agents to journal, track decisions, and extract generalized lessons to get smarter every day.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install clawmage-learning-loop
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install clawmage-learning-loop using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is Learning Loop?
Learning Loop is a sophisticated proactive learning engine designed to transform reactive AI agents into self-improving systems. Unlike standard memory setups that only update when a user provides a correction, this skill empowers agents to autonomously reflect on daily activities, manage a structured knowledge base, and synthesize long-term insights. It ensures that an agent doesn't just store data, but compounds knowledge over time through consistent reflection and review cycles.
By implementing a structured hierarchy for journaling, decision logging, and lesson extraction, Learning Loop provides a rigorous framework for an agent's intellectual growth. It focuses on generalization, asking how a specific solution can be applied to broader classes of problems, ensuring that the agent's performance scales alongside its experience. This is a core component of the Openclaw Skills framework for building truly capable digital entities.
Learning Loop Use Cases
- Setting up a new AI agent that needs to evolve and improve through continuous interaction.
- Reducing repetitive mistakes by implementing correction tracking and promotion rules.
- Creating a structured decision log to analyze agent accuracy across different categories like tool selection or architecture.
- Automating memory maintenance to keep context windows lean and relevant via hot, warm, and cold tiering.
- Establishing a daily reflection protocol to synthesize events and extract actionable lessons.
How Learning Loop Works
- The agent captures real-time events in a daily journal and logs significant decisions with clear reasoning and status tracking.
- When a user provides a correction, the agent logs the signal, increments a counter, and flags the pattern for nightly review.
- During the Nightly Reflection phase, the agent processes the day's logs, extracts new lessons, and files inbox items into the appropriate brain directory.
- Lessons undergo a lifecycle management process, moving from Active to Validated or Archived based on evidence and usage frequency.
- Staggered review chains (Monthly, Quarterly, Year-End) aggregate data to identify long-term patterns and maintain the health of the knowledge base.
- Memory tiering automatically moves inactive information from the primary memory file to archive storage to ensure high performance.
Learning Loop Setup
To initialize the Learning Loop structure within your workspace, ensure the following directory hierarchy is created:
mkdir -p brain/{journal,decisions,lessons,inbox,knowledge/{domains,people,concepts},projects,archive} memory/
Once the structure is in place, the agent will begin logging activities to brain/journal/YYYY-MM-DD.md and managing lessons within the brain/lessons/ directory. No external dependencies are required to start using this part of the Openclaw Skills suite.
Learning Loop Data Schema & Taxonomy
Learning Loop organizes data into a tiered 'brain' directory with the following taxonomy:
| Directory | Purpose | File Format |
|---|---|---|
journal/ |
Daily event logs and summaries | YYYY-MM-DD.md |
decisions/ |
Structured logs of consequential choices | DEC-NNN.md |
lessons/ |
Actionable insights with lifecycle status | LESSON-NNN.md |
knowledge/ |
Long-term mental models and domain data | Markdown files |
archive/ |
Stale or superseded information | Moved from active dirs |
Learning Loop Advanced Features
- Automatic Correction Promotion: Converts recurring user corrections into permanent rules after three occurrences.
- Quiet Day Protocol: Utilizes agent idle time for deep memory maintenance, cross-referencing, and documentation cleanup.
- Decision Accuracy Analytics: Tracks performance across categories (e.g., strategy vs. creative) to identify cognitive blind spots.
- Namespace Isolation: Implements a conflict resolution chain where project-specific knowledge overrides global domain rules.
- Memory Temperature Control: Strictly limits
MEMORY.mdto under 50 lines by demoting items unused for over 7 days to warm or cold storage.
SKILL.md
Loading
Related Openclaw Skills
Pretext Layout
A specialized tool for high-speed multiline text measurement and manual layout in browser environments that avoids expensive DOM reflows.
Facebook Graph API Workflows
An advanced operational skill for managing Facebook Page content and engagement via direct Graph API HTTPS requests.
Freelancer Toolkit
An AI-powered personal business manager that replaces manual time trackers and billing forms with simple conversational intelligence.
Novel to Storyboard Generator
A cinematic tool for Openclaw Skills that transforms literary prose into production-ready film storyboard scripts and shot lists.
Kiipu CLI
A professional command-line interface for managing Kiipu posts, authentication states, and system diagnostics.
Product Price Compare
An intelligent AI agent skill that automates cross-platform product price fetching, promotion parsing, and final cost calculation.
Featured*
